@ai-sdk/openai 2.0.0-canary.12 → 2.0.0-canary.14

package/dist/internal/index.mjs

@@ -248,7 +248,13 @@ var openaiProviderOptions = z.object({
   /**
    * Parameters for prediction mode.
    */
-  prediction: z.record(z.any()).optional()
+  prediction: z.record(z.any()).optional(),
+  /**
+   * Whether to use structured outputs.
+   *
+   * @default true
+   */
+  structuredOutputs: z.boolean().optional()
 });
 
 // src/openai-error.ts
@@ -331,20 +337,17 @@ function prepareTools({
 
 // src/openai-chat-language-model.ts
 var OpenAIChatLanguageModel = class {
-  constructor(modelId, settings, config) {
+  constructor(modelId, config) {
     this.specificationVersion = "v2";
+    this.supportedUrls = {
+      "image/*": [/^https?:\/\/.*$/]
+    };
     this.modelId = modelId;
-    this.settings = settings;
     this.config = config;
   }
   get provider() {
     return this.config.provider;
   }
-  async getSupportedUrls() {
-    return {
-      "image/*": [/^https?:\/\/.*$/]
-    };
-  }
   async getArgs({
     prompt,
     maxOutputTokens,
@@ -367,13 +370,14 @@ var OpenAIChatLanguageModel = class {
       providerOptions,
       schema: openaiProviderOptions
     })) != null ? _a : {};
+    const structuredOutputs = (_b = openaiOptions.structuredOutputs) != null ? _b : true;
     if (topK != null) {
       warnings.push({
         type: "unsupported-setting",
         setting: "topK"
       });
     }
-    if ((responseFormat == null ? void 0 : responseFormat.type) === "json" && responseFormat.schema != null && !this.settings.structuredOutputs) {
+    if ((responseFormat == null ? void 0 : responseFormat.type) === "json" && responseFormat.schema != null && !structuredOutputs) {
       warnings.push({
         type: "unsupported-setting",
         setting: "responseFormat",
@@ -402,12 +406,12 @@ var OpenAIChatLanguageModel = class {
       presence_penalty: presencePenalty,
       response_format: (responseFormat == null ? void 0 : responseFormat.type) === "json" ? (
         // TODO convert into provider option
-        this.settings.structuredOutputs && responseFormat.schema != null ? {
+        structuredOutputs && responseFormat.schema != null ? {
           type: "json_schema",
           json_schema: {
             schema: responseFormat.schema,
             strict: true,
-            name: (_b = responseFormat.name) != null ? _b : "response",
+            name: (_c = responseFormat.name) != null ? _c : "response",
             description: responseFormat.description
           }
         } : { type: "json_object" }
@@ -487,7 +491,7 @@ var OpenAIChatLanguageModel = class {
     } = prepareTools({
       tools,
       toolChoice,
-      structuredOutputs: (_c = this.settings.structuredOutputs) != null ? _c : false
+      structuredOutputs
     });
     return {
       args: {
@@ -847,11 +851,23 @@ var reasoningModels = {
   "o1-preview-2024-09-12": {
     systemMessageMode: "remove"
   },
+  o3: {
+    systemMessageMode: "developer"
+  },
+  "o3-2025-04-16": {
+    systemMessageMode: "developer"
+  },
   "o3-mini": {
     systemMessageMode: "developer"
   },
   "o3-mini-2025-01-31": {
     systemMessageMode: "developer"
+  },
+  "o4-mini": {
+    systemMessageMode: "developer"
+  },
+  "o4-mini-2025-04-16": {
+    systemMessageMode: "developer"
   }
 };
 
@@ -872,13 +888,9 @@ import {
 } from "@ai-sdk/provider";
 function convertToOpenAICompletionPrompt({
   prompt,
-  inputFormat,
   user = "user",
   assistant = "assistant"
 }) {
-  if (inputFormat === "prompt" && prompt.length === 1 && prompt[0].role === "user" && prompt[0].content.length === 1 && prompt[0].content[0].type === "text") {
-    return { prompt: prompt[0].content[0].text };
-  }
   let text = "";
   if (prompt[0].role === "system") {
     text += `${prompt[0].content}
@@ -984,6 +996,9 @@ var openaiCompletionProviderOptions = z4.object({
 var OpenAICompletionLanguageModel = class {
   constructor(modelId, config) {
     this.specificationVersion = "v2";
+    this.supportedUrls = {
+      // No URLs are supported for completion models.
+    };
     this.modelId = modelId;
     this.config = config;
   }
@@ -993,13 +1008,7 @@ var OpenAICompletionLanguageModel = class {
   get provider() {
     return this.config.provider;
   }
-  async getSupportedUrls() {
-    return {
-      // no supported urls for completion models
-    };
-  }
   async getArgs({
-    inputFormat,
     prompt,
     maxOutputTokens,
     temperature,
@@ -1043,7 +1052,7 @@ var OpenAICompletionLanguageModel = class {
         details: "JSON response format is not supported."
       });
     }
-    const { prompt: completionPrompt, stopSequences } = convertToOpenAICompletionPrompt({ prompt, inputFormat });
+    const { prompt: completionPrompt, stopSequences } = convertToOpenAICompletionPrompt({ prompt });
     const stop = [...stopSequences != null ? stopSequences : [], ...userStopSequences != null ? userStopSequences : []];
     return {
       args: {
@@ -1252,23 +1261,16 @@ var openaiEmbeddingProviderOptions = z6.object({
 
 // src/openai-embedding-model.ts
 var OpenAIEmbeddingModel = class {
-  constructor(modelId, settings, config) {
+  constructor(modelId, config) {
     this.specificationVersion = "v2";
+    this.maxEmbeddingsPerCall = 2048;
+    this.supportsParallelCalls = true;
     this.modelId = modelId;
-    this.settings = settings;
     this.config = config;
   }
   get provider() {
     return this.config.provider;
   }
-  get maxEmbeddingsPerCall() {
-    var _a;
-    return (_a = this.settings.maxEmbeddingsPerCall) != null ? _a : 2048;
-  }
-  get supportsParallelCalls() {
-    var _a;
-    return (_a = this.settings.supportsParallelCalls) != null ? _a : true;
-  }
   async doEmbed({
     values,
     headers,
@@ -1347,7 +1349,7 @@ var OpenAIImageModel = class {
     this.modelId = modelId;
     this.settings = settings;
     this.config = config;
-    this.specificationVersion = "v1";
+    this.specificationVersion = "v2";
   }
   get maxImagesPerCall() {
     var _a, _b;
@@ -1423,14 +1425,36 @@ import {
   parseProviderOptions as parseProviderOptions4,
   postFormDataToApi
 } from "@ai-sdk/provider-utils";
+import { z as z10 } from "zod";
+
+// src/openai-transcription-options.ts
 import { z as z9 } from "zod";
-var openAIProviderOptionsSchema = z9.object({
+var openAITranscriptionProviderOptions = z9.object({
+  /**
+   * Additional information to include in the transcription response.
+   */
   include: z9.array(z9.string()).nullish(),
+  /**
+   * The language of the input audio in ISO-639-1 format.
+   */
   language: z9.string().nullish(),
+  /**
+   * An optional text to guide the model's style or continue a previous audio segment.
+   */
   prompt: z9.string().nullish(),
-  temperature: z9.number().min(0).max(1).nullish().default(0),
-  timestampGranularities: z9.array(z9.enum(["word", "segment"])).nullish().default(["segment"])
+  /**
+   * The sampling temperature, between 0 and 1.
+   * @default 0
+   */
+  temperature: z9.number().min(0).max(1).default(0).nullish(),
+  /**
+   * The timestamp granularities to populate for this transcription.
+   * @default ['segment']
+   */
+  timestampGranularities: z9.array(z9.enum(["word", "segment"])).default(["segment"]).nullish()
 });
+
+// src/openai-transcription-model.ts
 var languageMap = {
   afrikaans: "af",
   arabic: "ar",
@@ -1504,12 +1528,11 @@ var OpenAITranscriptionModel = class {
     mediaType,
     providerOptions
   }) {
-    var _a, _b, _c, _d, _e;
     const warnings = [];
     const openAIOptions = await parseProviderOptions4({
       provider: "openai",
       providerOptions,
-      schema: openAIProviderOptionsSchema
+      schema: openAITranscriptionProviderOptions
     });
     const formData = new FormData();
     const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
@@ -1517,15 +1540,14 @@ var OpenAITranscriptionModel = class {
     formData.append("file", new File([blob], "audio", { type: mediaType }));
     if (openAIOptions) {
       const transcriptionModelOptions = {
-        include: (_a = openAIOptions.include) != null ? _a : void 0,
-        language: (_b = openAIOptions.language) != null ? _b : void 0,
-        prompt: (_c = openAIOptions.prompt) != null ? _c : void 0,
-        temperature: (_d = openAIOptions.temperature) != null ? _d : void 0,
-        timestamp_granularities: (_e = openAIOptions.timestampGranularities) != null ? _e : void 0
+        include: openAIOptions.include,
+        language: openAIOptions.language,
+        prompt: openAIOptions.prompt,
+        temperature: openAIOptions.temperature,
+        timestamp_granularities: openAIOptions.timestampGranularities
       };
-      for (const key in transcriptionModelOptions) {
-        const value = transcriptionModelOptions[key];
-        if (value !== void 0) {
+      for (const [key, value] of Object.entries(transcriptionModelOptions)) {
+        if (value != null) {
           formData.append(key, String(value));
         }
       }
@@ -1577,15 +1599,15 @@ var OpenAITranscriptionModel = class {
     };
   }
 };
-var openaiTranscriptionResponseSchema = z9.object({
-  text: z9.string(),
-  language: z9.string().nullish(),
-  duration: z9.number().nullish(),
-  words: z9.array(
-    z9.object({
-      word: z9.string(),
-      start: z9.number(),
-      end: z9.number()
+var openaiTranscriptionResponseSchema = z10.object({
+  text: z10.string(),
+  language: z10.string().nullish(),
+  duration: z10.number().nullish(),
+  words: z10.array(
+    z10.object({
+      word: z10.string(),
+      start: z10.number(),
+      end: z10.number()
     })
   ).nullish()
 });
@@ -1597,10 +1619,10 @@ import {
   parseProviderOptions as parseProviderOptions5,
   postJsonToApi as postJsonToApi5
 } from "@ai-sdk/provider-utils";
-import { z as z10 } from "zod";
-var OpenAIProviderOptionsSchema = z10.object({
-  instructions: z10.string().nullish(),
-  speed: z10.number().min(0.25).max(4).default(1).nullish()
+import { z as z11 } from "zod";
+var OpenAIProviderOptionsSchema = z11.object({
+  instructions: z11.string().nullish(),
+  speed: z11.number().min(0.25).max(4).default(1).nullish()
 });
 var OpenAISpeechModel = class {
   constructor(modelId, config) {
@@ -1703,7 +1725,7 @@ import {
   parseProviderOptions as parseProviderOptions6,
   postJsonToApi as postJsonToApi6
 } from "@ai-sdk/provider-utils";
-import { z as z11 } from "zod";
+import { z as z12 } from "zod";
 
 // src/responses/convert-to-openai-responses-messages.ts
 import {
@@ -1916,13 +1938,11 @@ function prepareResponsesTools({
 var OpenAIResponsesLanguageModel = class {
   constructor(modelId, config) {
     this.specificationVersion = "v2";
-    this.modelId = modelId;
-    this.config = config;
-  }
-  async getSupportedUrls() {
-    return {
+    this.supportedUrls = {
       "image/*": [/^https?:\/\/.*$/]
     };
+    this.modelId = modelId;
+    this.config = config;
   }
   get provider() {
     return this.config.provider;
@@ -2068,55 +2088,55 @@ var OpenAIResponsesLanguageModel = class {
       body,
       failedResponseHandler: openaiFailedResponseHandler,
       successfulResponseHandler: createJsonResponseHandler6(
-        z11.object({
-          id: z11.string(),
-          created_at: z11.number(),
-          model: z11.string(),
-          output: z11.array(
-            z11.discriminatedUnion("type", [
-              z11.object({
-                type: z11.literal("message"),
-                role: z11.literal("assistant"),
-                content: z11.array(
-                  z11.object({
-                    type: z11.literal("output_text"),
-                    text: z11.string(),
-                    annotations: z11.array(
-                      z11.object({
-                        type: z11.literal("url_citation"),
-                        start_index: z11.number(),
-                        end_index: z11.number(),
-                        url: z11.string(),
-                        title: z11.string()
+        z12.object({
+          id: z12.string(),
+          created_at: z12.number(),
+          model: z12.string(),
+          output: z12.array(
+            z12.discriminatedUnion("type", [
+              z12.object({
+                type: z12.literal("message"),
+                role: z12.literal("assistant"),
+                content: z12.array(
+                  z12.object({
+                    type: z12.literal("output_text"),
+                    text: z12.string(),
+                    annotations: z12.array(
+                      z12.object({
+                        type: z12.literal("url_citation"),
+                        start_index: z12.number(),
+                        end_index: z12.number(),
+                        url: z12.string(),
+                        title: z12.string()
                       })
                     )
                   })
                 )
               }),
-              z11.object({
-                type: z11.literal("function_call"),
-                call_id: z11.string(),
-                name: z11.string(),
-                arguments: z11.string()
+              z12.object({
+                type: z12.literal("function_call"),
+                call_id: z12.string(),
+                name: z12.string(),
+                arguments: z12.string()
               }),
-              z11.object({
-                type: z11.literal("web_search_call")
+              z12.object({
+                type: z12.literal("web_search_call")
               }),
-              z11.object({
-                type: z11.literal("computer_call")
+              z12.object({
+                type: z12.literal("computer_call")
               }),
-              z11.object({
-                type: z11.literal("reasoning"),
-                summary: z11.array(
-                  z11.object({
-                    type: z11.literal("summary_text"),
-                    text: z11.string()
+              z12.object({
+                type: z12.literal("reasoning"),
+                summary: z12.array(
+                  z12.object({
+                    type: z12.literal("summary_text"),
+                    text: z12.string()
                   })
                 )
               })
             ])
           ),
-          incomplete_details: z11.object({ reason: z11.string() }).nullable(),
+          incomplete_details: z12.object({ reason: z12.string() }).nullable(),
           usage: usageSchema
         })
       ),
@@ -2330,86 +2350,86 @@ var OpenAIResponsesLanguageModel = class {
     };
   }
 };
-var usageSchema = z11.object({
-  input_tokens: z11.number(),
-  input_tokens_details: z11.object({ cached_tokens: z11.number().nullish() }).nullish(),
-  output_tokens: z11.number(),
-  output_tokens_details: z11.object({ reasoning_tokens: z11.number().nullish() }).nullish()
+var usageSchema = z12.object({
+  input_tokens: z12.number(),
+  input_tokens_details: z12.object({ cached_tokens: z12.number().nullish() }).nullish(),
+  output_tokens: z12.number(),
+  output_tokens_details: z12.object({ reasoning_tokens: z12.number().nullish() }).nullish()
 });
-var textDeltaChunkSchema = z11.object({
-  type: z11.literal("response.output_text.delta"),
-  delta: z11.string()
+var textDeltaChunkSchema = z12.object({
+  type: z12.literal("response.output_text.delta"),
+  delta: z12.string()
 });
-var responseFinishedChunkSchema = z11.object({
-  type: z11.enum(["response.completed", "response.incomplete"]),
-  response: z11.object({
-    incomplete_details: z11.object({ reason: z11.string() }).nullish(),
+var responseFinishedChunkSchema = z12.object({
+  type: z12.enum(["response.completed", "response.incomplete"]),
+  response: z12.object({
+    incomplete_details: z12.object({ reason: z12.string() }).nullish(),
     usage: usageSchema
   })
 });
-var responseCreatedChunkSchema = z11.object({
-  type: z11.literal("response.created"),
-  response: z11.object({
-    id: z11.string(),
-    created_at: z11.number(),
-    model: z11.string()
+var responseCreatedChunkSchema = z12.object({
+  type: z12.literal("response.created"),
+  response: z12.object({
+    id: z12.string(),
+    created_at: z12.number(),
+    model: z12.string()
   })
 });
-var responseOutputItemDoneSchema = z11.object({
-  type: z11.literal("response.output_item.done"),
-  output_index: z11.number(),
-  item: z11.discriminatedUnion("type", [
-    z11.object({
-      type: z11.literal("message")
+var responseOutputItemDoneSchema = z12.object({
+  type: z12.literal("response.output_item.done"),
+  output_index: z12.number(),
+  item: z12.discriminatedUnion("type", [
+    z12.object({
+      type: z12.literal("message")
     }),
-    z11.object({
-      type: z11.literal("function_call"),
-      id: z11.string(),
-      call_id: z11.string(),
-      name: z11.string(),
-      arguments: z11.string(),
-      status: z11.literal("completed")
+    z12.object({
+      type: z12.literal("function_call"),
+      id: z12.string(),
+      call_id: z12.string(),
+      name: z12.string(),
+      arguments: z12.string(),
+      status: z12.literal("completed")
     })
   ])
 });
-var responseFunctionCallArgumentsDeltaSchema = z11.object({
-  type: z11.literal("response.function_call_arguments.delta"),
-  item_id: z11.string(),
-  output_index: z11.number(),
-  delta: z11.string()
+var responseFunctionCallArgumentsDeltaSchema = z12.object({
+  type: z12.literal("response.function_call_arguments.delta"),
+  item_id: z12.string(),
+  output_index: z12.number(),
+  delta: z12.string()
 });
-var responseOutputItemAddedSchema = z11.object({
-  type: z11.literal("response.output_item.added"),
-  output_index: z11.number(),
-  item: z11.discriminatedUnion("type", [
-    z11.object({
-      type: z11.literal("message")
+var responseOutputItemAddedSchema = z12.object({
+  type: z12.literal("response.output_item.added"),
+  output_index: z12.number(),
+  item: z12.discriminatedUnion("type", [
+    z12.object({
+      type: z12.literal("message")
     }),
-    z11.object({
-      type: z11.literal("function_call"),
-      id: z11.string(),
-      call_id: z11.string(),
-      name: z11.string(),
-      arguments: z11.string()
+    z12.object({
+      type: z12.literal("function_call"),
+      id: z12.string(),
+      call_id: z12.string(),
+      name: z12.string(),
+      arguments: z12.string()
     })
   ])
 });
-var responseAnnotationAddedSchema = z11.object({
-  type: z11.literal("response.output_text.annotation.added"),
-  annotation: z11.object({
-    type: z11.literal("url_citation"),
-    url: z11.string(),
-    title: z11.string()
+var responseAnnotationAddedSchema = z12.object({
+  type: z12.literal("response.output_text.annotation.added"),
+  annotation: z12.object({
+    type: z12.literal("url_citation"),
+    url: z12.string(),
+    title: z12.string()
   })
 });
-var responseReasoningSummaryTextDeltaSchema = z11.object({
-  type: z11.literal("response.reasoning_summary_text.delta"),
-  item_id: z11.string(),
-  output_index: z11.number(),
-  summary_index: z11.number(),
-  delta: z11.string()
+var responseReasoningSummaryTextDeltaSchema = z12.object({
+  type: z12.literal("response.reasoning_summary_text.delta"),
+  item_id: z12.string(),
+  output_index: z12.number(),
+  summary_index: z12.number(),
+  delta: z12.string()
 });
-var openaiResponsesChunkSchema = z11.union([
+var openaiResponsesChunkSchema = z12.union([
   textDeltaChunkSchema,
   responseFinishedChunkSchema,
   responseCreatedChunkSchema,
@@ -2418,7 +2438,7 @@ var openaiResponsesChunkSchema = z11.union([
   responseOutputItemAddedSchema,
   responseAnnotationAddedSchema,
   responseReasoningSummaryTextDeltaSchema,
-  z11.object({ type: z11.string() }).passthrough()
+  z12.object({ type: z12.string() }).passthrough()
   // fallback for unknown chunks
 ]);
 function isTextDeltaChunk(chunk) {
@@ -2466,16 +2486,16 @@ function getResponsesModelConfig(modelId) {
     requiredAutoTruncation: false
   };
 }
-var openaiResponsesProviderOptionsSchema = z11.object({
-  metadata: z11.any().nullish(),
-  parallelToolCalls: z11.boolean().nullish(),
-  previousResponseId: z11.string().nullish(),
-  store: z11.boolean().nullish(),
-  user: z11.string().nullish(),
-  reasoningEffort: z11.string().nullish(),
-  strictSchemas: z11.boolean().nullish(),
-  instructions: z11.string().nullish(),
-  reasoningSummary: z11.string().nullish()
+var openaiResponsesProviderOptionsSchema = z12.object({
+  metadata: z12.any().nullish(),
+  parallelToolCalls: z12.boolean().nullish(),
+  previousResponseId: z12.string().nullish(),
+  store: z12.boolean().nullish(),
+  user: z12.string().nullish(),
+  reasoningEffort: z12.string().nullish(),
+  strictSchemas: z12.boolean().nullish(),
+  instructions: z12.string().nullish(),
+  reasoningSummary: z12.string().nullish()
 });
 export {
   OpenAIChatLanguageModel,
@@ -2487,6 +2507,7 @@ export {
   OpenAITranscriptionModel,
   hasDefaultResponseFormat,
   modelMaxImagesPerCall,
+  openAITranscriptionProviderOptions,
   openaiCompletionProviderOptions,
   openaiEmbeddingProviderOptions,
   openaiProviderOptions